An inter-laboratory comparison exercise (IE) for the determination of PM mass concentration with the reference gravimetric method (EN 12341:2014) was organised for the first time at European level in 2015. Such an IE comprises the comparison between the samplers used by the various participants. Samplers shall therefore be co-located for several days so that the number of test samples is large enough for statistical analyses. Thanks to the great interest for this type of IE and to the commitment from the AQUILA Network members, 22 Laboratories plus the European Reference Laboratory for Air Pollution deployed their instruments at a single place in the premises of the Joint Research Centre in Ispra (Italy) for the same period of time (13 Feb. – 9 Apr. 2015). For almost two months, 24 samplers for PM10 and 22 for PM2.5 collected ambient aerosol samples on the filter types used by each participant. Gravimetric analyses of the filters before and after sampling were performed by the participants in their own laboratories following their own conditioning and filter handling procedures.
The assigned daily values for PM10 and PM2.5 were calculated as the Consensus Value from all participant results following the standard ISO 13528:2015. They ranged from 3.5 to 106 µg/m³ for PM10, and from 2.1 to 96 µg/m³ for PM2.5.
The scope of this inter-laboratory comparison was to assess the robustness of the measurement process and to determine the performance of the participants’ values. From the statistical analyses, conclusions could be drawn.
The reproducibility of the method was 14% for PM10 and 17% for PM2.5.
The Data Quality Objective of ±25% at the Daily Limit Value for PM10 (50µg/m³) was not reached for only 1.3% of all the participants’ results over the whole IE concentration range.
For PM10, out of 1245 reported daily values, 17 (1.4%) were found to be unsatisfactory with respect to the z’-score criterion (|z’|>3). Close to half of these values were produced by one single laboratory. For PM2.5, out of 1193 daily averages, only 5 (0.4%) showed an unsatisfactory result. Questionable results (3|z’|>2) amounted 4.7% and 3.8% of all PM10 and PM2.5 data, respectively.
The En-score, which takes into account also the uncertainties reported by the participants, exceeded the critical threshold (|En|>1) for 21.2% and 22.7% of the PM10 and PM2.5 reported data, respectively.
Between cooled and not-cooled samplers, the average relative differences (6% and 4% for PM10 and PM2.5, respectively) are significant at the 99.9% confidence level.
More than half of the samples were collected on quartz fibre filters for both aerosol size fractions. Compared to the assigned values, only PTFE filters led to significant biases (99.9% confidence level) of -4% and – 6% in PM10 and PM2.5, respectively.